2.3 创建你的第一个 Skill

🎯 本章目标

通过实战创建三个实用 Skill，从简单到复杂，让你完全掌握 Skill 开发流程。

实战 1：Hello Skill（5 分钟入门）

场景

创建一个简单的问候 Skill，当用户说"你好"时，按公司规范回复。

Step 1：创建文件夹

bash

# Personal Skills 位置
mkdir -p ~/.claude/skills/hello-skill
cd ~/.claude/skills/hello-skill

Step 2：创建 SKILL.md

markdown

---
name: hello-skill
description: Professional greeting skill. Use when user says hello or asks for a greeting.
---

# Hello Skill

## Purpose
Provide professional greetings following company etiquette.

## Usage
When user greets:
1. Respond with: "Hello! Welcome to Acme Corp. How may I assist you today?"
2. Use friendly but professional tone
3. Always end with offer to help

## Examples
- User: "Hi"
  → Response: "Hello! Welcome to Acme Corp. How may I assist you today?"

- User: "Good morning"
  → Response: "Good morning! Welcome to Acme Corp. How may I assist you today?"

Step 3：测试

你: "Hi Claude"
Claude: （自动加载 hello-skill）
        "Hello! Welcome to Acme Corp. How may I assist you today?"

关键学习点

✅ YAML Frontmatter - 正确格式 ✅ description - 包含触发条件"when user says hello" ✅ 清晰指令 - 具体回复模板 ✅ 示例 - 帮助 Claude 理解预期行为

实战 2：Markdown Report Skill（30 分钟进阶）

场景

创建标准化的 Markdown 报告，包含固定格式和样式。

Step 1：创建文件夹结构

bash

mkdir -p ~/.claude/skills/markdown-report
cd ~/.claude/skills/markdown-report
mkdir templates examples

Step 2：创建模板文件

templates/report_template.md

markdown

# {Title}

**Date**: {Date}
**Author**: {Author}

---

## Executive Summary

{Summary}

## Key Findings

1. {Finding 1}
2. {Finding 2}
3. {Finding 3}

## Recommendations

- {Recommendation 1}
- {Recommendation 2}

## Conclusion

{Conclusion}

---

*Generated by Acme Reporting System*

Step 3：创建 SKILL.md

markdown

---
name: markdown-report
description: Generate formatted Markdown reports with standard template. Use for creating reports, summaries, or documentation.
---

# Markdown Report Generator

## Purpose
Create professional Markdown reports following company template.

## Template Structure
Use the template in `templates/report_template.md` with these sections:
- Executive Summary
- Key Findings (3-5 points)
- Recommendations
- Conclusion

## Process
1. Ask user for:
   - Report title
   - Key findings
   - Recommendations
2. Load template from `templates/report_template.md`
3. Replace placeholders:
   - `{Title}` - Report title
   - `{Date}` - Current date
   - `{Author}` - User's name or "Acme Team"
   - `{Summary}` - Brief overview
   - `{Finding X}` - Key points
   - `{Recommendation X}` - Action items
4. Save as `Report_YYYY-MM-DD.md`

## Style Guidelines
- Use **bold** for emphasis
- Use bullet points for lists
- Keep Executive Summary under 200 words
- Each finding should be one clear sentence

## Example Output
See `examples/sample_report.md`

Step 4：创建示例文件

examples/sample_report.md

markdown

# Q4 Sales Analysis

**Date**: 2024-11-14
**Author**: Acme Team

---

## Executive Summary

Q4 sales exceeded targets by 15%, driven by strong performance in the enterprise segment. Customer retention improved to 94%, while new customer acquisition grew 22% year-over-year.

## Key Findings

1. Enterprise segment contributed 60% of total revenue growth
2. Customer retention rate improved from 89% to 94%
3. Average deal size increased by 18%

## Recommendations

- Increase investment in enterprise sales team
- Launch customer success program to maintain retention
- Develop upsell strategy for existing accounts

## Conclusion

Strong Q4 performance positions us well for next year's growth targets.

---

*Generated by Acme Reporting System*

Step 5：测试使用

你: "Create a Q4 sales report"

Claude: （加载 markdown-report Skill）
        "I'll help you create a Q4 sales report using our standard template.

         Please provide:
         1. Key findings (3-5 points)
         2. Main recommendations
         3. Brief summary"

你: [提供信息]

Claude: （生成报告）
        "Report created: Report_2024-11-14.md

         Preview:
         # Q4 Sales Analysis
         [完整内容...]"

关键学习点

✅ 模板文件 - 分离模板和逻辑 ✅ 占位符 - 使用 {Variable} 格式 ✅ 示例文件 - 提供参考 ✅ 清晰流程 - 步骤化指导

实战 3：CSV Data Analyzer（60 分钟高级）

场景

分析 CSV 文件并生成可视化报告，包含 Python 脚本执行。

Step 1：创建完整结构

bash

mkdir -p ~/.claude/skills/csv-analyzer
cd ~/.claude/skills/csv-analyzer
mkdir scripts templates output examples

Step 2：创建 Python 分析脚本

scripts/analyze.py

python

#!/usr/bin/env python3
import pandas as pd
import sys
import json

def analyze_csv(file_path):
    """Analyze CSV and return summary statistics"""
    df = pd.read_csv(file_path)

    analysis = {
        "rows": len(df),
        "columns": list(df.columns),
        "numeric_summary": df.describe().to_dict(),
        "missing_values": df.isnull().sum().to_dict(),
        "data_types": df.dtypes.astype(str).to_dict()
    }

    return analysis

if __name__ == "__main__":
    if len(sys.argv) < 2:
        print("Usage: python analyze.py <csv_file>")
        sys.exit(1)

    result = analyze_csv(sys.argv[1])
    print(json.dumps(result, indent=2))

Step 3：创建可视化脚本

scripts/visualize.py

python

#!/usr/bin/env python3
import pandas as pd
import matplotlib.pyplot as plt
import sys

def create_charts(csv_file, output_dir="output"):
    """Create basic charts from CSV data"""
    df = pd.read_csv(csv_file)

    # Get numeric columns
    numeric_cols = df.select_dtypes(include=['number']).columns

    # Create charts for first 3 numeric columns
    for i, col in enumerate(numeric_cols[:3]):
        plt.figure(figsize=(10, 6))
        df[col].plot(kind='hist', bins=20)
        plt.title(f'Distribution of {col}')
        plt.xlabel(col)
        plt.ylabel('Frequency')
        plt.savefig(f'{output_dir}/chart_{i+1}_{col}.png')
        plt.close()

    print(f"Charts saved to {output_dir}/")

if __name__ == "__main__":
    if len(sys.argv) < 2:
        print("Usage: python visualize.py <csv_file>")
        sys.exit(1)

    create_charts(sys.argv[1])

Step 4：创建 SKILL.md

markdown

---
name: csv-analyzer
description: Analyze CSV files and generate summary statistics with visualizations. Use when user needs data analysis or wants to explore CSV data.
allowed-tools:
  - Read
  - Bash
  - Write
---

# CSV Data Analyzer

## Purpose
Analyze CSV data files and provide insights with charts.

## Capabilities
- Summary statistics (mean, median, std, min, max)
- Data type detection
- Missing value analysis
- Distribution charts
- Correlation analysis

## Requirements
- Python 3.7+
- pandas
- matplotlib

Install dependencies:
```bash
pip install pandas matplotlib

Usage Process

1. Load CSV File

Ask user for CSV file path or let them provide the file.

2. Run Analysis

bash

python scripts/analyze.py data.csv

This outputs:

Row/column counts
Data types
Summary statistics
Missing values

3. Generate Charts

bash

python scripts/visualize.py data.csv

Creates distribution charts for numeric columns in output/ folder.

4. Generate Report

Combine analysis results into a markdown report:

Data overview
Key statistics
Charts (embedded as images)
Insights and recommendations

Output Format

Analysis Report: analysis_report_YYYY-MM-DD.md

Charts: output/chart_1_*.png, output/chart_2_*.png, etc.

Example Workflow

User: "Analyze my sales data"
Claude:
  1. Load sales.csv
  2. Run: python scripts/analyze.py sales.csv
  3. Run: python scripts/visualize.py sales.csv
  4. Generate report with findings
  5. Return: "Analysis complete! Found 3 insights..."

Template Report

See templates/analysis_report_template.md for report format.

Troubleshooting

Error: pandas not found → Run: pip install pandas matplotlib

Error: CSV parsing failed → Check file encoding (use UTF-8) → Verify delimiter (comma vs semicolon)

Examples

See examples/sample_analysis.md for a complete analysis example.


### Step 5：创建报告模板

`templates/analysis_report_template.md`

```markdown
# Data Analysis Report

**File**: {filename}
**Date**: {date}

## Data Overview
- **Rows**: {row_count}
- **Columns**: {column_count}
- **Data Types**: {data_types}

## Summary Statistics

{statistics_table}

## Missing Values

{missing_values_table}

## Key Insights

1. {insight_1}
2. {insight_2}
3. {insight_3}

## Visualizations

![Distribution Chart 1](../output/chart_1.png)
![Distribution Chart 2](../output/chart_2.png)

## Recommendations

- {recommendation_1}
- {recommendation_2}

---
*Generated by CSV Analyzer Skill*

Step 6：创建示例

examples/sample_data.csv

csv

date,sales,region,product
2024-01-01,1500,North,Widget A
2024-01-02,2300,South,Widget B
2024-01-03,1800,East,Widget A
2024-01-04,2100,West,Widget B
2024-01-05,1900,North,Widget C

examples/sample_analysis.md

markdown

# Data Analysis Report

**File**: sample_data.csv
**Date**: 2024-11-14

## Data Overview
- **Rows**: 5
- **Columns**: 4
- **Data Types**: date(object), sales(int64), region(object), product(object)

## Summary Statistics

| Metric | Sales |
|--------|-------|
| Mean   | 1920  |
| Median | 1900  |
| Std    | 291   |
| Min    | 1500  |
| Max    | 2300  |

## Missing Values

No missing values detected.

## Key Insights

1. Average sales: $1,920 per transaction
2. Sales range: $1,500 - $2,300 (15% variance)
3. Most common region: North (2 occurrences)

## Visualizations

![Sales Distribution](../output/chart_1_sales.png)

## Recommendations

- Investigate South region's high performance
- Consider promoting Widget B (highest sales)
- Maintain consistent sales in North region

---
*Generated by CSV Analyzer Skill*

Step 7：测试完整流程

你: "Analyze this sales CSV file"
   [上传 sales.csv]

Claude: （加载 csv-analyzer Skill）
        "I'll analyze your sales data using the CSV Analyzer skill.

         Running analysis..."

        [执行: python scripts/analyze.py sales.csv]

        "Analysis complete! Here's what I found:
         - 100 rows, 5 columns
         - Average sales: $2,340
         - 3 missing values in 'region' column

         Generating visualizations..."

        [执行: python scripts/visualize.py sales.csv]

        "Charts created! Generating report..."

        [创建 analysis_report_2024-11-14.md]

        "Report ready: analysis_report_2024-11-14.md

         Key insights:
         1. Sales peaked in Q3
         2. North region underperforming
         3. Strong correlation between price and sales"

关键学习点

✅ 执行代码 - Python 脚本集成 ✅ 多步骤流程 - 分析 → 可视化 → 报告 ✅ 错误处理 - Troubleshooting 部分 ✅ 依赖管理 - 明确 requirements ✅ 完整示例 - 端到端演示

🎓 Skill 开发最佳实践

1. 从简单开始

第一版: 基本功能
  ↓
第二版: 添加模板
  ↓
第三版: 添加脚本
  ↓
第四版: 完善文档

2. 测试驱动开发

markdown

## Test Cases

### Test 1: Basic Analysis
Input: simple.csv (5 rows, 3 columns)
Expected: Summary statistics + 2 charts

### Test 2: Large File
Input: large.csv (10000 rows)
Expected: Complete within 30 seconds

### Test 3: Missing Data
Input: incomplete.csv (20% missing)
Expected: Report missing value locations

3. 版本控制

bash

# 使用 Git 管理 Skill
cd ~/.claude/skills/csv-analyzer
git init
git add .
git commit -m "Initial CSV analyzer skill"

4. 文档优先

顺序：

写 README（给人看）
写 SKILL.md（给 Claude 看）
写代码（实现功能）
写示例（验证效果）

🐛 常见问题排查

问题 1：Claude 不使用我的 Skill

可能原因：

description 不够具体

markdown

# ❌ 太模糊
description: Data tool

# ✅ 具体明确
description: Analyze CSV files and generate statistics. Use when working with CSV data.

文件路径错误

bash

# 检查文件位置
ls ~/.claude/skills/your-skill/SKILL.md

# 应该存在

YAML 格式错误

yaml

# ❌ 错误
name:csv-analyzer  # 缺少空格

# ✅ 正确
name: csv-analyzer

问题 2：脚本执行失败

排查步骤：

bash

# 1. 检查脚本权限
chmod +x scripts/analyze.py

# 2. 测试脚本单独运行
python scripts/analyze.py test.csv

# 3. 检查依赖
pip list | grep pandas

# 4. 查看错误日志
claude --debug

问题 3：Skill 加载慢

优化方案：

markdown

# 分离大文件
large-skill/
├── SKILL.md              # 主文件（简洁）
├── detailed_guide.md     # 详细文档（按需加载）
└── examples/             # 示例（按需加载）
    └── 100_examples.md

📋 开发检查清单

创建 Skill 前：

[ ] 明确单一功能
[ ] 选择存储位置
[ ] 设计文件结构

编写 SKILL.md：

[ ] YAML 格式正确
[ ] description 包含触发词
[ ] 步骤清晰具体
[ ] 示例完整

测试验证：

[ ] 手动测试脚本
[ ] Claude 正确触发
[ ] 输出符合预期
[ ] 错误处理完善

🎯 进阶挑战

尝试创建这些 Skill：

Git Commit Message Generator
- 分析 git diff
- 生成规范的 commit message
- 遵循 Conventional Commits
Code Review Checklist
- 检查代码质量
- 应用团队 coding standards
- 生成 review 报告
API Documentation Generator
- 读取 OpenAPI spec
- 生成 Markdown 文档
- 包含示例代码

下一步： 2.4 进阶技巧和最佳实践 - 学习高级技巧和团队协作！

2.3 创建你的第一个 Skill ​

🎯 本章目标 ​

实战 1：Hello Skill（5 分钟入门） ​

场景 ​

Step 1：创建文件夹 ​

Step 2：创建 SKILL.md ​

Step 3：测试 ​

关键学习点 ​

实战 2：Markdown Report Skill（30 分钟进阶） ​

场景 ​

Step 1：创建文件夹结构 ​

Step 2：创建模板文件 ​

Step 3：创建 SKILL.md ​

Step 4：创建示例文件 ​

Step 5：测试使用 ​

关键学习点 ​

实战 3：CSV Data Analyzer（60 分钟高级） ​

场景 ​

Step 1：创建完整结构 ​

Step 2：创建 Python 分析脚本 ​

Step 3：创建可视化脚本 ​

Step 4：创建 SKILL.md ​

Usage Process ​

1. Load CSV File ​

2. Run Analysis ​

3. Generate Charts ​

4. Generate Report ​

Output Format ​

Example Workflow ​

Template Report ​

Troubleshooting ​

Examples ​

Step 6：创建示例 ​

Step 7：测试完整流程 ​

关键学习点 ​

🎓 Skill 开发最佳实践 ​

1. 从简单开始 ​

2. 测试驱动开发 ​

3. 版本控制 ​

4. 文档优先 ​

🐛 常见问题排查 ​

问题 1：Claude 不使用我的 Skill ​

问题 2：脚本执行失败 ​

问题 3：Skill 加载慢 ​

📋 开发检查清单 ​

🎯 进阶挑战 ​

2.3 创建你的第一个 Skill

🎯 本章目标

实战 1：Hello Skill（5 分钟入门）

场景

Step 1：创建文件夹

Step 2：创建 SKILL.md

Step 3：测试

关键学习点

实战 2：Markdown Report Skill（30 分钟进阶）

场景

Step 1：创建文件夹结构

Step 2：创建模板文件

Step 3：创建 SKILL.md

Step 4：创建示例文件

Step 5：测试使用

关键学习点

实战 3：CSV Data Analyzer（60 分钟高级）

场景

Step 1：创建完整结构

Step 2：创建 Python 分析脚本

Step 3：创建可视化脚本

Step 4：创建 SKILL.md

Usage Process

1. Load CSV File

2. Run Analysis

3. Generate Charts

4. Generate Report

Output Format

Example Workflow

Template Report

Troubleshooting

Examples

Step 6：创建示例

Step 7：测试完整流程

关键学习点

🎓 Skill 开发最佳实践

1. 从简单开始

2. 测试驱动开发

3. 版本控制

4. 文档优先

🐛 常见问题排查

问题 1：Claude 不使用我的 Skill

问题 2：脚本执行失败

问题 3：Skill 加载慢

📋 开发检查清单

🎯 进阶挑战